[sterlind]

That one would be interesting to try with AlphaZero. Reinforcement learning systems tend to have trouble with imperfect information.